1: \documentclass[prl,twocolumn,showpacs,amsmath,amssymb]{revtex4}
2: %\documentclass[prl,galley,showpacs,amsmath,amssymb]{revtex4}
3:
4: % Some other (several out of many) possibilities
5: %\documentclass[preprint,aps]{revtex4}
6: %\documentclass[preprint,aps,draft]{revtex4}
7: %\documentclass[prb]{revtex4}% Physical Review B
8:
9: \usepackage{graphicx}% Include figure files
10: \usepackage{dcolumn}% Align table columns on decimal point
11: \usepackage{bm}% bold math
12:
13: %%%
14: \def\tmo{TbMnO$_3$}
15: \def\phm#1{\phantom{#1}}
16: \def\be{\begin{equation}}
17: \def\ee{\end{equation}}
18: %%%
19:
20:
21: \begin{document}
22:
23: %\title{First-principles study of magnetoelectric coupling in \tmo}
24: \title{First-principles study of improper ferroelectricity in \tmo}
25:
26: \author{Andrei Malashevich}
27: \email{andreim@physics.rutgers.edu}
28: \author{David Vanderbilt}
29: \affiliation{
30: Department of Physics \& Astronomy, Rutgers University,
31: Piscataway, NJ 08854-8019, USA
32: }
33:
34: \date{\today}
35:
36: \begin{abstract}
37: We have carried out a first-principles theoretical study of the
38: magnetically induced polarization in orthorhombic \tmo, a prototypical
39: material in which a cycloidal spin structure generates an
40: electric polarization via the spin-orbit interaction. We compute
41: both the electronic and the lattice-mediated contributions
42: to the polarization and find that the latter is strongly dominant.
43: We analyze the spin-orbit induced forces and lattice displacements
44: from both atomic and mode-decomposition viewpoints,
45: and show that a simple model based on nearest Mn--Mn neighbor
46: Dzyaloshinskii-Moriya interactions is not able to account fully
47: for the results. The direction and magnitude of our computed polarization
48: are in good agreement with experiment.
49: \end{abstract}
50:
51: \pacs{75.80.+q,77.80.-e}% PACS, the Physics and Astronomy
52: % Classification Scheme.
53:
54: \maketitle
55:
56: Multiferroic and magnetoelectric materials are currently under
57: intensive study, both from the point of view of fundamental
58: materials physics and because of their potential application in
59: novel technological devices such as memories, sensors, and
60: transducers \cite{fiebig}. An interesting class of systems are
61: ``improper ferroelectrics''
62: in which the magnetic order induces an electric polarization
63: in a material that is otherwise structurally centrosymmetric. This
64: can happen in collinear spin systems if the spin pattern
65: breaks inversion symmetry; examples include
66: $R$Mn$_2$O$_5$ ($R$=Tb, Y, etc.) \cite{chapon:2004,chapon:2006},
67: the E phase of orthorhombic $R$MnO$_3$ ($R$=Ho, Tb)
68: \cite{munoz:2001,Sergienko:2006a}, \tmo\ in magnetic field
69: \cite{aliouane:2006}, and Ca$_3$Co$_{2-x}$Mn$_x$O$_6$
70: \cite{choi:2008}.
71: In such cases, symmetric exchange interactions can
72: explain the development of polar order, i.e., via exchange
73: striction effects \cite{Sergienko:2006a,cheong-mostovoy}.
74: Here we shall be concerned with another class of systems in which
75: a non-collinear cycloidal spin structure induces an electric
76: polarization via the spin-orbit (SO) interaction
77: \cite{Kimura:2007}; examples include Ni$_3$V$_2$O$_8$ \cite{lawes:2005},
78: CuFeO$_2$ \cite{Kimura:2006}, and orthorhombic $R$MnO$_3$
79: ($R$=Tb, Dy, Ho, etc.) \cite{kimura,Kenzelmann:2005,Yamasaki-both}.
80: While such spiral spin structures and the resulting polarizations
81: usually appear only at low temperature, these systems are of special
82: interest because the coupling between magnetic and electric degrees
83: of freedom is so profound.
84:
85: Theoretically, symmetry analysis and phenomenological models
86: have clarified the circumstances under which a spiral spin structure
87: can give rise to an electric polarization
88: \cite{Mostovoy:2006,Radaelli:2007,harris},
89: but are not well suited to identifying the dominant microscopic
90: mechanism responsible for the polar order. Microscopic models
91: have been introduced and used to discuss possible mechanisms, which
92: can be divided broadly into two categories. The first are purely
93: electronic mechanisms, in which the SO interaction (SOI) modifies the
94: hybridization of electronic orbitals in such a way as to shift the
95: center of charge
96: \cite{Katsura:2005,Bruno:2005,jia:2006,jia:2007}.
97: The second are ``lattice mechanisms'' involving magnetically-induced
98: ionic displacements, usually discussed in terms of Dzyaloshinskii-Moriya
99: (DM) interactions \cite{Sergienko:2006,harris,Hu:2007}.
100: However, it is generally difficult to estimate the magnitudes (and
101: even signs) of these two kinds of contributions from the models.
102: Because experiments have not been sensitive enough to resolve the
103: tiny SO-induced atomic displacements, a central unanswered question is
104: whether electronic or lattice mechanisms are dominant.
105: It is also unclear whether the
106: pattern of displacements should follow the predictions of
107: a simple DM model, or whether more complicated interactions enter
108: the picture.
109:
110: First-principles calculations can play an invaluable role in resolving
111: such questions.
112: Density-functional theory (DFT) has proven to be an extremely useful
113: tool in understanding the physics of ordinary ferroelectric materials,
114: and has already been used to make important contributions to the
115: understanding of improper magnetic ferroelectrics of the collinear-spin
116: type \cite{hill_m,wang,picozzi:2007} and in the spiral magnetic materials LiCu$_2$O$_2$
117: and LiCuVO$_4$ \cite{xiang_whangbo}. However, to our knowledge such methods
118: have not previously been applied to study the cycloidal magnetic states
119: in the orthorhombic perovskite system.
120:
121: In this Letter, we choose orthorhombic \tmo\ as a paradigmatic
122: system of the cycloidal type, and carry out a detailed
123: first-principles study of the mechanisms by which the cycloidal
124: magnetic order gives rise to the polarization. We demonstrate
125: that DFT calculations predict the appearance of electric polarization
126: in the cycloidal state via two mechanisms, a purely electronic one
127: and a lattice-mediated one, and that the latter is strongly dominant.
128: Nevertheless, we find that the DM theory alone cannot explain
129: the computed pattern of atomic displacements, suggesting that
130: other interactions not included in simple DM models play an
131: important role.
132:
133: Ferroelectricity coupled to spiral magnetic order was discovered in
134: the orthorhombic perovskite \tmo\ (space group {\it Pbnm})
135: \cite{kimura} with a 20-atom cell containing four formula units.
136: The Mn$^{3+}$ ions form a collinear sinusoidal spin wave at
137: temperatures between $\sim$27\,K and $\sim$41\,K, but below
138: $\sim$27\,K a cycloidal spin wave forms with incommensurate
139: wavevector $k_{\rm s}\sim0.28$ along $b$, and a polarization
140: simultaneously appears along $c$ \cite{kimura,Kimura:2007}.
141:
142: In this work, electronic-structure calculations are
143: carried out using a plane-wave pseudopotential approach to
144: DFT as implemented within the VASP code \cite{kresse-vasp}
145: using PAW potentials\cite{Blochl,Kresse:1999}.
146: (The Tb potential does not have $f$ electrons in the valence.)
147: We use the local-density approximation (Ceperley-Alder\cite{Ceperley}
148: with Vosko-Wilk-Nusair correlation\cite{voskown})
149: with on-site Coulomb interactions
150: (LDA+U). We let $U$=1\,eV to make the band gap close to the experimental
151: value of $\sim0.5$\,eV \cite{Cui:2005}.
152: Since the first-principles calculations require periodic boundary
153: conditions, we study a system in which the cycloidal
154: spin structure is commensurate with $k_{\rm s}=1/3$, making a supercell
155: that is tripled in $b$-direction (60 atoms per cell) as shown in
156: Fig.\,\ref{fig:rums}. This is close enough to the
157: experimental $k_{\rm s}$ of 0.28 that our results can be expected
158: to be qualitatively and semiquantitatively meaningful. A
159: 3$\times$1$\times$2 $k$-point sampling is used.
160: The plane-wave energy cutoff was 500 eV, and
161: the electric polarization was calculated using the Berry-phase
162: approach \cite{King-Smith}.
163:
164: \begin{figure}
165: \includegraphics[height=1.7in]{tmo_new.eps}
166: \caption{\label{fig:rums}(Color online)
167: Sketch of $a\times3b\times c/2$ orthorhombic cell of \tmo\ ({\it Pbnm}
168: space group) showing Mn$^{3+}$ sites with a cycloidal magnetic structure.}
169: \end{figure}
170:
171: A structural relaxation was performed for the 60-atom supercell
172: without SO. The resulting structural coordinates, reported in Table
173: \ref{tab:struct}, are in quite satisfactory agreement with experiment.
174: Our computed magnetic moments have magnitudes $3.4\,\mu_{\rm B}$
175: and rotate in the $b-c$ plane as one scans along $b$, and this
176: result hardly changes when SO is turned on. Experimentally, the
177: magnetic moments on Mn$^{3+}$ sites are found to form an elliptical
178: spiral with amplitudes $m_b=3.9\,\mu_{\rm B}$ and $m_c=2.8\,\mu_{\rm B}$
179: \cite{Kenzelmann:2005}. The origin of this ellipticity, which is
180: apparently not captured by our calculation, deserves further investigation.
181:
182: \begin{table}[t]
183: \caption{\label{tab:struct} Experimental (Ref.~\onlinecite{alonso})
184: and theoretical (this work) structural parameters for orthorhombic
185: \tmo.}
186: \begin{ruledtabular}
187: \begin{tabular}{clccc}
188: & & & Experiment & Theory\\
189: \hline
190: \multicolumn{2}{l}{Lattice vectors}
191: & $a$ (\AA) & 5.293 & 5.195 \\
192: & & $b$ (\AA) & 5.838 & 5.758 \\
193: & & $c$ (\AA) & 7.403 & 7.308 \\
194: Tb & 4c($x$ $y$ 1/4) & $x$ & 0.983 & 0.979 \\
195: & & $y$ & 0.082 & 0.084 \\
196: Mn & 4b(1/2 0 0) & & & \\
197: O1 & 4c($x$ $y$ 1/4) & $x$ & 0.104 & 0.107 \\
198: & & $y$ & 0.467 & 0.469 \\
199: O2 & 8d($x$ $y$ $z$) & $x$ & 0.704 & 0.699 \\
200: & & $y$ & 0.326 & 0.320 \\
201: & & $z$ & 0.051 & 0.052
202: \end{tabular}
203: \end{ruledtabular}
204: \end{table}
205:
206: Next, we confirmed that the Berry-phase polarization vanishes in the
207: absence of SO. When the SO is turned on,
208: one expects the broken inversion symmetry in the spin sector
209: to be communicated to the spatial (charge) degrees of freedom, and
210: indeed our calculations confirm this. Focusing first on the purely
211: electronic mechanism, we froze the lattice vectors and ionic positions
212: while turning on the SOI and computed the polarization to be
213: $P=32\,\mu{\rm C}/{\rm m}^2$ along the $c$ direction\cite{uncert}
214: (and zero along directions $a$ and $b$).
215: This is much smaller than the experimental
216: value of $\sim600\,\mu{\rm C}/{\rm m}^2$ \cite{kimura}, suggesting that
217: the lattice mechanism must be dominant. We thus turned our attention
218: next to the role of ionic displacements.
219:
220: As a first step, we computed the Hellmann-Feynman forces appearing
221: on the ions as the SO is turned on (with the structural coordinates
222: still clamped). Using the 20-atom cell of the $Pbnm$ space group to
223: label the forces, we found that:
224: (i) 64.8\% of the forces belong to the symmetry-preserving A$_1(\Gamma)$
225: irreducible representation (irrep);
226: (ii) 28.7\% of the forces belong to the B$_{\rm 3u}(\Gamma)$
227: irrep, i.e., zone-center infrared-active (IR-active) modes with
228: dynamical dipoles along $c$;
229: and (iii) 6.5\% correspond to forces at $k_y=\pm2\pi/3b$, i.e.,
230: those that would be responsible for lowering the translational
231: symmetry of the 20-atom cell.
232: The remaining contributions are found to vanish to
233: numerical precision. Since the only forces
234: that generate an electric polarization in linear order are those
235: of type (ii), we henceforth focus our attention on the zone-center
236: IR-active B$_{\rm 3u}$ modes.
237:
238: Eight Wyckoff coordinates contribute to the B$_{\rm 3u}$ irrep in the
239: 20-atom {\it Pbnm} structure:
240: three associated with Mn atoms in Wyckoff position 4b,
241: one each for Tb and O1 atoms in Wyckoff position 4c,
242: and three for the O2 atoms in Wyckoff position 8d \cite{bcs:sam}.
243: To find the polarization induced by the SOI,
244: we need to deduce the displacements resulting from
245: these forces. To this end we calculated the 8$\times$8
246: force-constant matrix $\Phi_{ij}=-\partial{f_i}/\partial{u_j}$
247: connecting the B$_{\rm 3u}$ forces and displacements at linear order using
248: finite-difference methods (i.e., we shift each Wyckoff coordinate
249: by $\sim$$10^{-3}$\,\AA\ and calculate the resulting forces).
250: For computational convenience, the determination of $\Phi_{ij}$ was
251: done in the absence of SO. (We checked a few elements and found
252: hardly any difference in the presence of SO \cite{explan-so}.)
253: Furthermore, we identify the linear combination
254: $t_j$ of the eight coordinates corresponding to a uniform shift of the
255: crystal along the $c$ direction (acoustic mode) and make sure to
256: project this combination out of the computed forces and
257: displacements. Thus, the displacements are predicted from
258: $u_i=-\sum^8_{j=1}\tilde\Phi^{-1}_{ij}f_j$,
259: where $\tilde\Phi^{-1}$ is a pseudo-inverse \cite{Xifan} having the
260: properties that
261: $\tilde\Phi^{-1} \Phi u=u$ for any $u$ orthogonal to $t$.
262:
263: The computed forces and predicted atomic displacements in the
264: IR-active B$_{\rm 3u}$ sector for \tmo\ are presented in the first
265: three columns of Table \ref{tab:wyck_modes}. We calculated the
266: B$_{\rm 3u}(\Gamma)$ forces again in the presence of the predicted
267: displacements and found that 99\% of these forces
268: were eliminated, thus justifying our approach
269: \cite{explan-fc}. The distorted crystal structure is now
270: non-centrosymmetric (space group $Pna2_1$), and the
271: calculated Berry-phase polarization
272: (including both electronic and ionic contributions) is
273: $P=-467\,\mu{\rm C}/{\rm m}^2$. This is more than an order of
274: magnitude larger than the value of 32\,$\mu$C/m$^2$ computed from
275: purely electronic mechanisms, and comparable in magnitude with the
276: experimental value of $\sim-$600$\,\mu$C/m$^2$ \cite{Yamasaki-both}.
277: Note that the sign of the ionic contribution is different from the
278: purely electronic contribution, making the sign of the total polarization
279: consistent with experiment.
280: We thus arrive at the first major conclusion of our paper, namely,
281: that the lattice contributions dominate strongly
282: and have approximately the right magnitude
283: to explain the experimentally observed polarization.
284: Purely electronic mechanisms, e.g.\ as described by the
285: Katsura-Nagaosa-Balatsky (KNB) model\cite{Katsura:2005} or related
286: work \cite{Bruno:2005,jia:2006,jia:2007}, are not sufficient to
287: describe the ferroelectricity in \tmo.
288:
289: \begin{table}[t]
290: \caption{\label{tab:wyck_modes} Forces, displacements, effective charges
291: and contributions to ionic polarization from IR-active B$_{\rm 3u}$ modes.}
292: \begin{ruledtabular}
293: \begin{tabular}{lrrrr}
294: ~~Wyckoff & $F$~~~ & $\Delta u$~~ & $Z^{*}$ & $\Delta{P}$~~~ \\
295: coordinate & (meV/\AA) & (m\AA) & ($e$) & ($\mu$C/m$^2$) \\
296: \hline
297: 1 (Tb 4c, $z$) & 0.434 & $-$0.171 & 7.468 & $-$94 \\
298: 2 (O1 4c, $z$) & 2.264 & $-$0.162 & $-$6.815 & 81 \\
299: 3 (Mn 4b, $x$) & $-$7.039 & $-$0.315 & 0.567 & $-$13 \\
300: 4 (Mn 4b, $z$) & $-$8.933 & $-$0.453 & 7.460 & $-$248 \\
301: 5 (Mn 4b, $y$) & $-$2.940 & $-$0.012 & 0.553 & $-$0.5 \\
302: 6 (O2 8d, $x$) & 5.058 & 0.540 & 0.077 & 3 \\
303: 7 (O2 8d, $y$) & 3.573 & 0.934 & 0.227 & 16 \\
304: 8 (O2 8d, $z$) & 4.409 & 0.556 & $-$5.738 & $-$234
305: \end{tabular}
306: \end{ruledtabular}
307: \end{table}
308:
309: We focus henceforth on the lattice-mediated contribution to the
310: polarization.
311: To understand which modes contribute most strongly and
312: to test whether simple models (based, e.g., on DM interactions)
313: can predict the forces and displacements, we performed
314: a further analysis as follows. For each
315: B$_{\rm 3u}$ Wyckoff coordinate, we computed the corresponding effective
316: charge $Z^*$ (derivative of polarization with respect to the Wyckoff
317: coordinate) by augmenting the same finite-difference calculations
318: used earlier with Berry-phase polarization calculations. We then
319: multiplied effective charges and displacements to get the contribution
320: $\Delta P$ made by each Wyckoff coordinate.
321: The results are given in the last two columns of Table \ref{tab:wyck_modes}.
322: The total ionic polarization (sum of the last column) is
323: $-$489\,$\mu$C/m$^2$, in good agreement with the change of polarization
324: in going from the centrosymmetric to the distorted structure as
325: reported earlier ($-$467$-$32=$-$499\,$\mu$C/m$^2$).
326: Not surprisingly, the $Z^*$ values are much larger for the four Wyckoff
327: coordinates involving displacements along $z$, and the corresponding
328: $\Delta P$ values give by far the largest contribution.
329:
330: In a simple DM model \cite{Sergienko:2006a}, one considers interactions
331: involving nearest-neighbor triplets of ions along Mn--O--Mn bonds.
332: Such a picture leads to the expectation that a force
333: $\gamma\,\hat{{\bf e}}_{nn'}\times {\bf S}_n\times {\bf S}_{n'}$
334: should appear on the central
335: O ion (where $\gamma$ is a coupling, ${\bf S}_n$ and ${\bf S}_{n'}$ are
336: neighboring Mn spins, and
337: $\hat{{\bf e}}_{nn'}$ is the unit vector connecting them),
338: with an equal and opposite force shared between the Mn ions.
339: The cycloidal spin structure of \tmo\ is such that O1 ions should be
340: unaffected, and O2 ions should all feel the same force along
341: $c$, with Mn ions feeling an opposite force. Thus, within
342: such a model one would expect only the 4th and 8th rows of
343: Table \ref{tab:wyck_modes} to give significant contributions.
344: Inspecting the Table, we see that
345: this is rather far from being the case; one does find substantial
346: oppositely-directed forces and displacements for the (Mn 4b, $z$)
347: and (O2 8d, $z$) coordinates, but the forces and displacements
348: are comparable for some other coordinates. The
349: contributions to $\Delta P$ are more strongly
350: dominated by these two coordinates, but only because
351: the $Z^*$ values are so much larger for displacements
352: along $\hat{z}$.
353: Interestingly, the Tb displacements are responsible for $\sim$20\%
354: of the lattice polarization, despite being uninvolved
355: in the usual picture of Mn--O--Mn couplings.
356: Thus we conclude that the simple DM model, and
357: indeed any model that considers only nearest-neighbor Mn--Mn spin
358: interactions, is unable to provide a detailed description
359: of the ferroelectricity in \tmo.
360:
361: \begin{table}[t]
362: \caption{\label{tab:eig_modes} Forces, displacements, effective charges
363: and contributions to ionic polarization from modes corresponding
364: to eigenvectors of the force-constant matrix.}
365: \begin{ruledtabular}
366: \begin{tabular}{rrrrr}
367: Eigenvalue & $F$~~~ & $\Delta{u}$~~ & $Z^{*}$ & $\Delta{P}$~~~ \\
368: (eV/\AA$^2$) & (meV/\AA) & (m\AA) & (e) & ($\mu$C/m$^2$) \\
369: \hline
370: 34.86 & 6.349 & 0.182 & $-$8.111 & $-$108 \\
371: 31.37 & $-$8.259 & $-$0.263 & 4.565 & $-$88 \\
372: 13.32 & 2.783 & 0.209 & 8.132 & 125 \\
373: 11.87 & 7.614 & 0.641 & $-$2.127 & $-$100 \\
374: 10.35 & 1.684 & 0.163 & $-$3.406 & $-$41 \\
375: 6.99 & $-$2.784 & $-$0.398 & 4.316 & $-$126 \\
376: 3.85 & 4.032 & 1.047 & $-$1.991 & $-$153
377: \end{tabular}
378: \end{ruledtabular}
379: \end{table}
380:
381: As an alternative way of analyzing the SO-induced B$_{\rm 3u}$
382: forces and displacements, we transform into the basis of
383: eigenvectors of the 8$\times$8 force-constant matrix $\Phi$. The
384: results are presented in Table \ref{tab:eig_modes}.
385: (Components on the translational zero mode, not shown in
386: the Table, were numerically negligible.)
387: We find that the forces are rather widely distributed among these
388: ``modes.'' Not surprisingly, the displacements are largest for
389: the softest mode, but other modes also show substantial displacements.
390: Moreover, the mode dynamical charges are much larger for some of
391: the harder modes, with the result that the contributions to
392: the polarization are of comparable magnitude for almost all of the
393: B$_{\rm 3u}$ modes. Thus, it is clear that the physics of the
394: electric polarization in TbMnO$_3$ is not dominated by a single
395: softest mode.
396:
397: To investigate the role of spin-orbit effects located on the various atomic
398: sites, we performed additional calculations of the
399: forces (and, via the force-constant matrix, the resulting
400: lattice polarizations). We first turned
401: the SOI off on all sites other than Tb, then Mn, then
402: O, finding contributions to the polarization of
403: $-11\,\mu$C/m$^2$, $-447\,\mu$C/m$^2$ and $-8\,\mu$C/m$^2$,
404: respectively. (The sum is in fairly good agreement with the total
405: ionic polarization of $-$489\,$\mu$C/m$^2$; this and other tests
406: confirm that the results are roughly linear in SO coupling.)
407: This confirms that the SOI on the Mn sites is responsible for almost
408: all of the lattice-mediated contribution to the polarization, even
409: though some of these forces appear on other sites,
410: even including Tb (see Table \ref{tab:wyck_modes}).
411:
412: Up to now our analysis has been done using the 60-atom
413: supercell, whose cycloidal wavevector ($k_{\rm s}$=1/3)
414: approximates the experimental one ($k_{\rm s}$=$0.28$).
415: To investigate how the results depend
416: on $k_s$, we have also carried calculations of the SO-induced
417: forces for a 40-atom supercell ($k_{\rm s}$=$1/2$). In order to
418: isolate the effects of the spin-wave period, we calculated the
419: forces at the same strucural coordinates as for the 60-atom
420: structure, and again projected into the B$_{\rm 3u}(\Gamma)$ irrep.
421: We find that the pattern of B$_{\rm 3u}$ forces matches almost
422: exactly that of the 60-atom cell (with the angle between the two
423: eight-dimensional vectors being $\sim$3$^\circ$), although the
424: magnitude is larger for the 40-atom cell. From the simple DM model
425: one expects
426: ${\bf P} \propto {\bf e}_{n,n'}\times({\bf S}_n\times{\bf S}_{n'})$
427: \cite{Kimura:2007} where the spins are neighbors on the zig-zag chains
428: propagating along $b$ (see Fig.~\ref{fig:rums}). The angle between
429: the spins is 90$^\circ$ and 60$^\circ$ for the 40-atom and 60-atom
430: supercells respectively, giving an expected ratio
431: of $2:\sqrt{3}$=1.15. In the long-wavelength
432: limit one expects ${\bf P} \propto |k_{\rm s}|$, for a ratio of 1.5.
433: Strangely, our computed ratio for the 40- and 60-atom supercells is
434: about 1.48,
435: closer to the continuum expression in spite of the rather short
436: wavelength of the spiral. This may be
437: another indication that further-neighbor Mn--Mn
438: interactions, or other effects not captured by a
439: nearest-neighbor DM model, play an important role.
440: From the forces we predict a lattice
441: polarization of $-754\,\mu$C/m$^2$ for $k_{\rm s}$=1/2;
442: extrapolating through our $k_{\rm s}$=1/3 result, we estimate a
443: lattice polarization of $-410\,\mu$C/m$^2$ for $k_{\rm s}$=0.28.
444:
445: In summary, we have used first-principles methods to compute the
446: electronic and lattice contributions to the spin-orbit induced
447: electric polarization in the cycloidal-spin compound \tmo, and
448: find the lattice contribution to be strongly dominant.
449: We provide a detailed analysis of the lattice contributions coming from
450: individual sites and individual modes and roughly characterize
451: the dependence of the lattice polarization on spin-wave period.
452: Our result for the polarization agrees in sign and compares fairly
453: well in magnitude with the experimentally measured value of
454: $\sim-$600$\,\mu$C/m$^2$ \cite{Yamasaki-both}.
455:
456: The authors wish to thank Morrel H.\ Cohen and Craig J.\ Fennie
457: for fruitful discussions.
458: This work was supported by NSF Grant No. DMR-0549198.
459:
460: {\it Note:} While finalizing this manuscript, we learned of
461: closely related work by H.J.\ Xiang {\it et al.} \cite{Xiang:2008}
462:
463: \bibliography{tmo}
464:
465:
466: \end{document}
467: